Recently, with the development of the communication and the computer technology and the improvement of the storage
technology and the capability of the digital image equipment, more and more image resources are given to us than ever.
And thus the solution of how to locate the proper image quickly and accurately is wanted.The early method is to set up a
key word for searching in the database, but now the method has become very difficult when we search much more
picture that we need. In order to overcome the limitation of the traditional searching method, content based image
retrieval technology was aroused. Now, it is a hot research subject.Color image retrieval is the important part of it. Color
is the most important feature for color image retrieval. Three key questions on how to make use of the color
characteristic are discussed in the paper: the expression of color, the abstraction of color characteristic and the
measurement of likeness based on color. On the basis, the extraction technology of the color histogram characteristic is
especially discussed. Considering the advantages and disadvantages of the overall histogram and the partition histogram,
a new method based the partition-overall histogram is proposed. The basic thought of it is to divide the image space
according to a certain strategy, and then calculate color histogram of each block as the color feature of this block. Users
choose the blocks that contain important space information, confirming the right value. The system calculates the
distance between the corresponding blocks that users choosed. Other blocks merge into part overall histograms again,
and the distance should be calculated. Then accumulate all the distance as the real distance between two pictures. The
partition-overall histogram comprehensive utilizes advantages of two methods above, by choosing blocks makes the
feature contain more spatial information which can improve performance; the distances between partition-overall
histogram make rotating and translation does not change. The HSV color space is used to show color characteristic of
image, which is suitable to the visual characteristic of human. Taking advance of human's feeling to color, it quantifies
color sector with unequal interval, and get characteristic vector. Finally, it matches the similarity of image with the
algorithm of the histogram intersection and the partition-overall histogram. Users can choose a demonstration image to
show inquired vision require, and also can adjust several right value through the relevance-feedback method to obtain the
best result of search.An image retrieval system based on these approaches is presented. The result of the experiments
shows that the image retrieval based on partition-overall histogram can keep the space distribution information while
abstracting color feature efficiently, and it is superior to the normal color histograms in precision rate while researching.
The query precision rate is more than 95%. In addition, the efficient block expression will lower the complicate degree of
the images to be searched, and thus the searching efficiency will be increased. The image retrieval algorithms based on
the partition-overall histogram proposed in the paper is efficient and effective.
|